SOD1 (Copper/Zinc Superoxide Dismutase) Deficiency Drives Amyloid β Protein Oligomerization and Memory Loss in Mouse Model of Alzheimer Disease*

Oxidative stress is closely linked to the pathogenesis of neurodegeneration. Soluble amyloid β (Aβ) oligomers cause cognitive impairment and synaptic dysfunction in Alzheimer disease (AD). However, the relationship between oligomers, oxidative stress, and their localization during disease progression is uncertain. Our previous study demonstrated that mice deficient in cytoplasmic copper/zinc superoxide dismutase (CuZn-SOD, SOD1) have features of drusen formation, a hallmark of age-related macular degeneration (Imamura, Y., Noda, S., Hashizume, K., Shinoda, K., Yamaguchi, M., Uchiyama, S., Shimizu, T., Mizushima, Y., Shirasawa, T., and Tsubota, K. (2006) Proc. Natl. Acad. Sci. U.S.A. 103, 11282–11287). Amyloid assembly has been implicated as a common mechanism of plaque and drusen formation. Here, we show that Sod1 deficiency in an amyloid precursor protein-overexpressing mouse model (AD mouse, Tg2576) accelerated Aβ oligomerization and memory impairment as compared with control AD mouse and that these phenomena were basically mediated by oxidative damage. The increased plaque and neuronal inflammation were accompanied by the generation of Nϵ-carboxymethyl lysine in advanced glycation end products, a rapid marker of oxidative damage, induced by Sod1 gene-dependent reduction. The Sod1 deletion also caused Tau phosphorylation and the lower levels of synaptophysin. Furthermore, the levels of SOD1 were significantly decreased in human AD patients rather than non-AD age-matched individuals, but mitochondrial SOD (Mn-SOD, SOD2) and extracellular SOD (CuZn-SOD, SOD3) were not. These findings suggest that cytoplasmic superoxide radical plays a critical role in the pathogenesis of AD. Activation of Sod1 may be a therapeutic strategy for the inhibition of AD progression.

Recently, there have been reports on the involvement of mitochondrial oxidative stress in the pathogenesis of AD; transgenic mouse models of AD were crossed with Sod2 ϩ/Ϫ , resulting in increased plaque formation (17), accelerated behavioral deficits (18), and the hyperphosphorylation of Tau (19), but no effects on A␤ oligomerization were reported. Information on the contribution of SOD1 to A␤ oligomer formation and the localization is, therefore, required to further elucidate the role of cytoplasmic superoxide in the mechanism of AD.
To achieve this we generated an AD model mouse lacking Sod1 and analyzed it for AD-like pathology. This report shows that cytoplasmic SOD reduction induced A␤ oligomerization, causing cognitive impairment, and that neuronal dysfunction is mediated by oxidative damage of brain tissues. Consistent with the animal studies, the levels of SOD1, but not those of SOD2 or SOD3, were significantly decreased in the brains of human AD subjects compared with non-AD individuals, thereby highlighting a potential causative role for SOD1-mediated A␤ oligomerization in the pathogenesis of AD.

EXPERIMENTAL PROCEDURES
Mice-Tg2576 (20) expressing the human APP (hAPP, C57BL6/SJL background) possessing the Swedish mutation shows the "early" (cognitive impairment) and "late" phenotypes (plaque formation) of AD (21). We utilized Tg2576 (Taconic) because it is possible to evaluate each effect of SOD1 deletion on the cognitive function or amyloid depositions. Sod1 Ϫ/Ϫ mice (The Jackson Laboratory) were backcrossed with C57BL/ 6NCrSlc mice for 5 or 6 generations. To generate hAPP/ Sod1 ϩ/Ϫ mice, we bred hAPP/Sod1 ϩ/ϩ mice with Sod1 Ϫ/Ϫ mice. Studies were conducted on sibling offspring from hAPP/ Sod1 ϩ/Ϫ ϫ Sod1 ϩ/Ϫ matings, giving hAPP/Sod1 Ϫ/Ϫ mice together with the corresponding littermate controls (hAPP/ Sod1 ϩ/Ϫ , hAPP/Sod1 ϩ/ϩ , Sod1 Ϫ/Ϫ , Sod1 ϩ/Ϫ , and Sod1 ϩ/ϩ ). All experiments were performed at two age points; a young age (6 -8 months old) and old age groups (15-17 months old), in which groups consisted of sex-balanced females and males. Mice were genotyped by PCR using genomic DNA isolated from the tail tip as reported previously (13). The animals were housed under a 12-h light/dark cycle and with ad libitum access to food and water. The mice were maintained and studied according to protocols approved by the Animal Care Committee of the Tokyo Metropolitan Institute of Gerontology. All experiments were performed by examiners blinded to the genotypes of the mice.
Human Brain-The frontal lobes of the brains of AD (6 female, 4 male) and non-AD (5 female, 5 male) individuals (Table 1) were used in the experiment with written informed consent obtained from the patients' families, and the experiment was approved by the Ethics Committee of Tokyo Metro-politan Institute of Gerontology and Tokyo Metropolitan Geriatric Hospital. The National Institute on Aging-Reagan criteria (modified) were adopted for the diagnosis of AD (22). The normal controls were defined using clinical documentation of unimpaired cognition as well as minimal senile changes consisting of Braak's neurofibrillary tangle stage equal to or less than II, senile plaque stage equal to or less than A, and a lack of any vascular, inflammatory, or traumatic changes or tumors.
Tissue Preparation and Western Blotting-The procedure was based on the method of previous works (23)(24)(25)(26)(27). In brief, human brain tissue (0.1-0.2 g) was homogenized in 10 volumes (w/v) of 50 mM TBS (Tris-HCl buffer (pH 7.6) containing 150 mM NaCl, a mixture of protease and phosphatase inhibitors (Complete TM ; Roche Diagnostics) supplemented with 0.7 g/ml pepstatin A and 1 mM phenylmethylsulfonyl fluoride). The animal brain was then removed, split into two hemispheres, and one-half was frozen on liquid nitrogen. The other hemisphere was fixed in 4% paraformaldehyde and embedded in paraffin for immunohistochemical study. Indeed, there are the differences of distribution for plaque depositions in several hAPP transgenic mice. In preliminary experiments, we investigated the variance of plaque distribution using several lines (Tg2576 (20), J20 (28), and PS2Tg2576 (27)); Tg2576 and PS2Tg2576 had almost no preference in the plaque areas, whereas the plaque deposits was frequently found in the hippocampus of J20, as Mucke et al. (28) also previously described. Therefore, we carried out the biochemical analysis using the whole brain lysates of mice crossed with Tg2576. The mouse brain (ϳ250 mg) was homogenized in 3 volumes (w/v) of TBS buffer as described above, and homogenates were centrifuged at 186,000 ϫ g for 30 min at 4°C using an Optima TL ultracentrifuge and a TLA100.4 rotor to give the supernatant (TBSsoluble) and pellet (TBS-insoluble) fractions. The pellet was then dissolved by sonication in 70% formic acid containing a mixture of protease inhibitors. The solubilized pellet was centrifuged at 186,000 ϫ g for 30 min at 4°C, after which the supernatant was neutralized with 1 M Tris base of pH 11 (1:20, v:v) as an insoluble fraction (formic acid-soluble). The total protein concentration of the brain was determined using the DC protein assay (Bio-Rad).
Dot Blotting-Two microliters of TBS-soluble fractions (2 g/l) were applied to a nitrocellulose membrane (0.2-m

SOD1-deficient AD Mice Show Early Memory Loss
pore size, Bio-Rad) basically according to the protocol developed by Kayed et al. (29). After being blocked, the membrane was incubated with anti-A␤ oligomer (A11, 0.1 g/ml, Invitrogen) overnight at 4°C before being incubated with the secondary antibody (1 h). As implied previously (29), TBS containing 0.01% Tween 20 was used as a washing buffer to reduce the interference with the detection of oligomers by higher concentrations of detergent. Development was performed with enhanced chemiluminescence and quantified using LAS-3000 (Fujifilm).
Morris Water Maze-Memory impairment was assessed using the Morris water maze test, as described previously (25)(26)(27). The water maze pool (Muromachi Kikai, Tokyo), diameter 120 cm, contained opaque water (20°C) with a platform (10 cm in diameter) submerged 2 cm below the surface. The hidden platform task took 4 (6 -8-month group) or 7 (15-17-month group) days (2 sessions per day, 3 h apart) during which 2 trials were performed each day (15 min apart). The platform location remained constant, and the entry points were changed semirandomly between trials. Twenty-four hours after the last day of the hidden platform task, a 1-min probe trial was carried out without the platform. The entry point for the probe trials was in the quadrant opposite the target quadrant. Memory retention was evaluated by the amount of time spent in the correct quadrant where the escape platform was located in the hidden platform trial. Performance was monitored with the CompACT VAS/DV video-tracking system.
Y maze-Exploratory behavior was tested using the Y maze test, as described previously (24 -26). The Y maze apparatus (Muromachi Kikai) was made of gray plastic walls, 12 cm high consisting of three compartments (40 ϫ 2 cm) connected with 2 ϫ 2-cm passages. The mice were placed into one of the three arms of the maze and allowed to explore the two open arms for 5 min, during which the third arm remained closed (training trial). After a 1-h interval, the closed arm was opened, and the mice were allowed to explore all three arms for 5 min (test trial). An arm entry was recorded when all four paws entered the compartment. After testing each mouse, the maze was thoroughly cleaned to standardize odors. Performance was monitored with the CompACT VAS/DV video-tracking system.
Immunohistochemistry-The procedure was basically adopted based on the method of the previous works (23)(24)(25)(26)(27). In brief, the sections (5 m) were deparaffinized, rehydrated, and washed in phosphate-buffered saline (PBS) before being treated briefly with formic acid in the case of A␤ staining. After incubation in 3% hydrogen peroxide in methanol to prevent endogenous peroxidation, the sections were blocked with 10% normal goat serum in PBS before being incubated with an anti-A␤ antibody (6E10 and 4G8) (1 g/ml; Signet), an anti-ionized calcium binding adaptor molecule 1 (Iba-1) antibody (1 g/ml; Wako), to detect activated microglia or an anti-glial fibrillary acid protein (GFAP) antibody (1 g/ml; Sigma) to detect activated astrocytes overnight at 4°C. The sections were incubated with biotinylated secondary antibody for 30 min. Immunoreactivity was visualized using an ABC Elite kit according to the manufacturer's protocol. The sections were counterstained with hematoxylin, and densitometric quantification of the percentage area was measured using Leica QWin V3 image software.
Thioflavin-S Staining-After deparaffinization of the sections and washing with PBS, the slides were immersed for 5 min in 0.25% potassium permanganate solution followed by 5 min in 1% potassium metabisulfate, 1% oxalic acid solution. The treated slides were placed in a filtered 0.02% thioflavin-S solution for 8 min before being subjected to florescence photomicrography in aqueous mounting medium.
Determination of Oxidative DNA Damage-Nuclear DNA was isolated, and 8-hydroxydeoxyguanosine (8-OHdG) was measured using an electrochemical detection-HPLC system as described previously (30). The 8-OHdG content was expressed as the molar ratio of 8-OHdG to 10 7 of dG. The amount of dG was calculated from the absorption at 260 nm in the same sample.
Quantification of Glutathione-The glutathione (GSH) content was determined using a total glutathione quantification kit (Dojindo) according to the manufacturer's instructions. The concentration of GSH in the sample was measured from absorbance at 412 nm. 20-l samples (2 g/l) of the TBSsoluble fraction were employed.
Measurement of Red Blood Cells-Twenty microliters of fresh blood obtained from the tail vein were subjected to automatic hematological analysis (Celltac, MEK-6358, Nihon Kohden, Tokyo, Japan) after a 100 times dilution with isotonic buffer (Isotonac, MEK-510, Nihon Kohden). Human blood (MEK-3DN, Nihon Kohden) was used as the standard.
Statistical Analyses-All data are presented as the means Ϯ S.E., and the differences were analyzed with one-way analysis of variance (ANOVA) followed by Bonferroni's test. p values Ͻ0.05 were considered significant.

Ablation of Sod1 Allele Reduces SOD1 Levels of hAPP Mice
-To determine whether SOD1 contributes to the A␤-dependent pathology of AD in vivo, we generated hAPP/Sod1 Ϫ/Ϫ by crossing hAPP/Sod1 ϩ/Ϫ mice with Sod1 ϩ/Ϫ mice, giving six types of littermates: hAPP/Sod1 ϩ/ϩ , hAPP/Sod1 ϩ/Ϫ , hAPP/ Sod1 Ϫ/Ϫ , Sod1 ϩ/ϩ , Sod1 ϩ/Ϫ , and Sod1 Ϫ/Ϫ . These groups did not differ with regard to the background strain. We confirmed the approximate half-reduction of SOD1 levels in hAPP/ Sod1 ϩ/Ϫ mice at 6 -8 and 15-17 months old in a gene dose-dependent manner using immunoblot analysis (Fig. 1, A-C). The overexpression of hAPP also did not significantly affect the amounts of SOD1 in hAPP/Sod1 ϩ/ϩ . To examine the effect of hAPP overexpression on the anemia phenotype of the Sod1 Ϫ/Ϫ mice (31), the red blood cells of these mice were counted, and no significant change was observed in either age group by overexpression of hAPP ( Fig. 1 D and E).
Sod1 Deletion Accelerates Loss of Spatial Learning and Memory of hAPP Mice-During the acquisition period of Morris maze, mice were trained to search for the hidden platform for 4 days. At 6 -8 months of age, all mice quickly achieved the goal except for hAPP/Sod1 Ϫ/Ϫ mice, which took significantly longer to master this task ( Fig. 2A). Probe trials, in which the platform was removed and mice were given 1 min to find the missing platform, were performed 1 day after the hidden platform task. As shown in Fig. 2B, only hAPP/Sod1 Ϫ/Ϫ mice failed to find the target location. Estimation of time spent in the target quadrant in the probe trial showed no apparent recognition by hAPP/ Sod1 Ϫ/Ϫ mice of the target zone (Fig. 2C), consistent with the results of the hidden trial test ( Fig. 2A). There were no differences in swimming speed among all mice investigated (data not shown), indicating that motor functions were normal. Furthermore, no drusen deposits were observed in 6 -8-month-old mice knocking out Sod1 (data not shown), as described previously (11), meaning that visual functions were also normal in young mice.
At 15-17 months of age, hAPP/Sod1 ϩ/ϩ , hAPP/Sod1 ϩ/Ϫ , and hAPP/Sod1 Ϫ/Ϫ mice took longer to reach the goal than control Sod1 ϩ/ϩ mice (supplemental Fig. S1A). In the probe trial all mice possessing the hAPP transgene showed poor memory of the goal zone (supplemental Fig. S1B), and no pref-erence for the target area was found in these mice (supplemental Fig. S1C). All mice tested in this age group swam at a similar speed in the probe trial (supplemental Fig. S1D). A visible platform trial was carried out before the hidden trial to ensure that the significant difference was not due to visual or motor dysfunctions (data not shown). Although we confirmed the presence of drusens in 15-17-month-old mice without Sod1 (hAPP/Sod1 Ϫ/Ϫ and Sod1 Ϫ/Ϫ , supplemental Fig. S2), consistent with our previous studies (11), hAPP/Sod1 ϩ/Ϫ showed the significant memory loss compared with Sod1 ϩ/ϩ . These results suggest that cytoplasmic SOD deletion significantly accelerates the A␤-dependent learning and memory deficits of hAPP/ Sod1 ϩ/ϩ mice in an age-dependent manner.
Sod1 Reduction Accelerates Behavioral Abnormality of hAPP Mice-The Y maze depends on the natural tendency of rodents to explore new environments. As a preference for the novel arm reflects locomotor activity, novel arm entries and the time spent in the novel arm by hAPP mice in the test trial were recorded after the training trial. Young hAPP/Sod1 ϩ/ϩ mice (age 6 -8 months) visited the novel arm as often as non-Tg mice, whereas both hAPP/Sod1 Ϫ/Ϫ and hAPP/Sod1 ϩ/Ϫ mice entered the novel arm less frequently than non-Tg mice (Fig.  2D). hAPP/Sod1 Ϫ/Ϫ and hAPP/Sod1 ϩ/Ϫ mice also showed poor recognition of the novel arm as determined by the percentage of time spent there (Fig. 2E). Although haplo-deficiency of Sod1 induced behavioral impairment in hAPP/Sod1 ϩ/ϩ mice in the Y maze, these alterations were generally coincident with the learning and memory deficits observed in the Morris water maze test (Fig. 2, A-C), indicating that cytoplasmic SOD plays a role in behavioral function.
Sod1 Ablation Alters Neuronal Dysfunction and Tau Abnormality of hAPP Mice-Neuronal degeneration in animal models is closely associated with memory and behavioral impairments. We examined the levels of synaptophysin as a neuronal marker that is expressed on presynaptic vesicles by Western blotting. Of the young mice in the 6 -8-month-old group, only hAPP/Sod1 Ϫ/Ϫ showed measurable decreases in synaptophysin levels within the TBS-soluble fraction (Fig. 3A). In older animals all hAPP-overexpressing mice showed decreased synaptophysin (Fig. 3B). There were almost no changes on the levels of synaptophysin by Sod1 reduction both in young and old non-Tg groups (supplemental Fig. S3, A and B).
Phosphorylation of Tau and its accumulation in the neurofibrillary tangles are characteristics of the pathogenesis of AD (3). The phosphorylation of Tau at Ser-396 is closely related to its abnormal polymerization (32). We observed a significant increase in the phosphorylation of Tau in the 6 -8-month-old hAPP/Sod1 Ϫ/Ϫ mice, whereas no alternation of its total amount compared with hAPP/Sod1 ϩ/ϩ mice (Fig. 3, C and E). Similarly, Sod1 reduction induced the higher phosphorylation of Tau in 15-17-month-old hAPP/Sod1 Ϫ/Ϫ , although almost no increases in total Tau among three 15-17-month-old mice groups were obtained (Fig. 3, D and F). On the other hand, Sod1 down-regulation did not significantly affect the total Tau and phosphorylation of Tau both in young and old non-Tg groups (supplemental Fig. S3, C-F). The difference in the ratio of phosphorylated Tau to total Tau among groups reached significance in both age Tg groups (Fig. 3, G and H), whereas that in both age shown are representative Western blots of SOD1 in the brains of mice of the indicated genotypes and age groups along with ␤-actin as an internal standard. B and C, shown is densitometric quantification of immunoblot signals for SOD1 normalized to ␤-actin of mice at the age of 6 -8 months (n ϭ 4ϳ5 per genotype) and 15-17 months (n ϭ 4ϳ5 per genotype). *, p Ͻ 0.05; **, p Ͻ 0.01 versus either Sod1 ϩ/ϩ , mean Ϯ S.E. D and E, data are presented of red blood cells (RBC) of the mice at the age of 6 -8 months (n ϭ 13ϳ17 per genotype) and 15-17 months (n ϭ 5ϳ17 per genotype). *, p Ͻ 0.05, mean Ϯ S.E. non-Tg groups did not (supplemental Fig. S3, G and H). These findings suggest that impairment of memory or behavioral function induced by SOD1 decline is associated with presynaptic protein loss and/or Tau phosphorylation.
Sod1 Reduction Exacerbates Plaque Formation of hAPP Mice-To investigate the effect of Sod1 deficiency on plaque load in the brain, immunohistochemistry was performed using anti-A␤ antibodies (6E10 and 4G8). Robust formation of senile plaques in old hAPP/Sod1 Ϫ/Ϫ mice (age, 15-17 months) was found (Fig.  4A). The area occupied by plaques in the neocortex and hippocampus was quantified by densitometric analysis. The plaque load of hAPP/Sod1 Ϫ/Ϫ and hAPP/Sod1 ϩ/Ϫ mice was markedly increased by about 2-fold as compared with that of hAPP/Sod1 ϩ/ϩ mice (Fig. 4B). To confirm the results of immunohistochemical analysis, the concentration of brain A␤ in the TBS-insoluble fraction was measured by ELISA. In agreement with the calculation of plaque area (Fig. 4B), Fig. 4C shows that the levels of A␤42 (286%) and A␤40 (223%) in hAPP/Sod1 Ϫ/Ϫ mice were about twice that observed in hAPP/Sod1 ϩ/ϩ mice, whereas hAPP/Sod1 ϩ/Ϫ mice also showed a significant increase in A␤42 (262%). These results indicate that in older animals the deletion of SOD1 increases the insoluble amyloid burden.
In contrast, the onset of small deposits was observed in 6 -8-month-old hAPP/Sod1 Ϫ/Ϫ mice, which was confirmed by thioflavin-S fluorescence, whereas hAPP/Sod1 ϩ/ϩ mice showed no plaques (Fig. 4D); however, the total levels of A␤42 and A␤40 in the TBS-insoluble fraction were not altered by SOD1 changes in this age group as determined by ELISA (Fig. 4, E and F).
Sod1 Reduction Alters Microglia and Astrocytes of hAPP Mice-The infiltration of activated microglia and astrocytes is linked to plaque formation (33). We performed immunohistochemistry to visualize reactive microglia and astrocytes using Iba-1 and GFAP antibodies, respectively. As shown in Fig. 5A, the immunostaining of Iba-1 in the hippocampus demonstrated intense activation of microglia in old hAPP/Sod1 Ϫ/Ϫ mice. Potent elevation of GFAP immunoreactivity was also seen in old hAPP/Sod1 Ϫ/Ϫ mice, demonstrating the activation of astrocytes (Fig. 5B). Quantitative analysis of the stained surface area showed significant increases in the numbers of reactive microglia (235%) and astrocytes (228%) in hAPP/Sod1 Ϫ/Ϫ mice compared with those in hAPP/Sod1 ϩ/ϩ mice, with over twice the up-regulation observed in hAPP/Sod1 Ϫ/Ϫ mice (Fig. 5, C and  D). Although the levels of both Iba-1 and GFAP levels slightly increased by Sod1 reduction in the old non-Tg group, these changes did not reach the significance (p ϭ 0.14, 0.28, respectively). These data imply that the augmented immunoreactivity induced by SOD1 down-regulation is consistent with robust plaque formation in Tg groups. Very weak activation of microglia or astrocytes was observed in the 6 -8-month-old mouse  DECEMBER 30, 2011 • VOLUME 286 • NUMBER 52 groups (supplemental Fig. S4), supporting the correlation of the neuroinflammation with plaque formation.

SOD1-deficient AD Mice Show Early Memory Loss
Sod1 Ablation Modulates Oxidative Damage of hAPP Mice-We sought to determine whether the accelerated AD-like phenotype caused by SOD1 reduction correlates with the oxidative damage to protein or DNA in brain tissues. Protein oxidation occurs after the reaction of amino acid residues (e.g. lysine, cysteine, and arginine) with carbohydrates, leading to the formation of advanced glycation end-products. The CML in advanced glycation end-product is thought to be one of the most rapid markers of oxidative damage (34). To examine the effect of the Sod1 deletion on oxidative stress in hAPP mice, Western blotting of CML derivatization was performed. The blots showed mainly three bands immunoreactive with anti-CML antibody at 67, 50, and 45 kDa, generally consistent with previous studies (35). Interestingly, in the 6 -8-month-old mice groups, the intensities of these CML bands of only hAPP/ Sod1 Ϫ/Ϫ mice were enhanced, whereas no effects of Sod1 reduction were seen in young non-Tg groups (Fig. 6, A and B). On the other hand, the generation of CML in both groups of 15-17-month-old non-Tg and Tg mice was accelerated in a gene dose-dependent manner of Sod1 (Fig. 6, A and C). Importantly, the CML levels of old hAPP/Sod1 ϩ/ϩ mice were higher than Sod1 ϩ/ϩ (Fig. 6C), as mentioned in our previous studies using another line (J20) of AD mouse model (25).
Oxidation of nuclear DNA is also an established marker of AD (30). The effect on nuclear DNA isolated from the brain was investigated using HPLC measurements (Fig. 6, D and E). There was a significant increase in the ratio of 8-OHdG to dG in 15-17-month-old hAPP/Sod1 Ϫ/Ϫ mice compared with hAPP/ Sod1 ϩ/ϩ and hAPP/Sod1 ϩ/Ϫ mice (Fig. 6E). hAPP/Sod1 ϩ/Ϫ and hAPP/Sod1 ϩ/ϩ mice also showed an elevated ratio of 8-OHdG over hAPP/Sod1 ϩ/ϩ and Sod1 ϩ/ϩ mice, respectively. On the other hand, there was no significant increase by SOD1 reduction in the younger groups, although a slight increase was found in non-Tg groups (Fig. 6D). The different vulnerability between proteins and DNA imply that the primary target may be proteins rather than DNA in AD progression and the relevance of CML derivatization with AD.
Sod1 Deletion Accelerates A␤ Oligomer Formation of hAPP Mice-To clarify the potential effect of SOD1 reduction on A␤ assembly, we examined the levels of soluble A␤42 and A␤40 in the TBS-soluble fraction using ELISA. No significant increase was observed in either monomeric A␤42 or A␤40 levels in young mice groups (Fig. 7, A and C), respectively, whereas the soluble monomer amounts for A␤42 and A␤40 were augmented in old hAPP/Sod1 Ϫ/Ϫ mice (Fig. 7, B and D). Interestingly, the dot-blotting using A11 oligomer antibody developed by Glabe and co-workers (29) revealed that the formation of A11-reactive oligomers was accelerated by Sod1 deletion both in young and old hAPP/Sod1 Ϫ/Ϫ (Fig. 7, E and F), especially more extensively in young mice groups. No A11-reactive oligomers were found in three non-Tg groups for both ages (Fig. 7,  E and F). To evaluate the size of increased oligomers in Fig. 7, E and F, Western blotting for A␤ using 82E1 antibody against the N terminus of A␤ was performed (Fig. 7, G and H). The ϳ60-kDa bands as well as ϳ30-kDa bands were significantly increased by Sod1 reduction, indicating that one of the oligomeric species might be the 12-mer or 6-mer of A␤. Similar tendency was observed using 6E10 (data not shown). On the other hand, non-Tg groups did not show any A␤ bands for both ages (supplemental Fig. S5, A and B). These results did not contradict the previous studies that one of the A11-reactive oligomers could be ϳ55 kDa (36). The low amount of low molecular weight oligomers could be due to the procedure to make water-soluble homogenates using TBS without detergent. In addition, the monomer band as a major A␤ was found in 82E1 Western blotting using TBS-insoluble fractions of three Tg groups, and there was no significant difference between three Tg groups for both ages (supplemental Fig. S5, C and D). It was also confirmed that any A␤ bands were not seen in three non-Tg groups for both ages (supplemental Fig. S5, E and F). These results did not contradict the previous report by Tomiyama et al. (37), in particular, that the monomer and low molecular weight oligomers in TBS without detergent could be hardly observed even in their Osaka mutant-Tg mice (APP E693⌬ -Tg mice) favoring oligomer formation.
Moreover, using an ELISA specific for A␤ oligomers developed by Selkoe and co-workers (38), in which the same N-terminal A␤ antibody (82E1) was used both for antigen capture and detection, the amount of soluble oligomers was significantly increased (14%) in 6 -8-month-old hAPP/Sod1 Ϫ/Ϫ mice compared with hAPP/Sod1 ϩ/ϩ mice (supplemental Fig. S5G), albeit at a lower level, and these changes were inversely correlated with the decline of cognitive function shown in Fig. 2. At 15-17 months of age (supplemental Fig. S5H), the deficiency of SOD1 increased oligomer formation in hAPP/Sod1 ϩ/ϩ mice (426%). The oligomer in old hAPP/Sod1 Ϫ/Ϫ mice was increased as compared with hAPP/Sod1 ϩ/ϩ mice in addition to their elevated levels of both A␤42 and A␤40 monomer (Fig. 7, B and D), strongly implying the specific elevation of oligomerization in the 6 -8-month-old group. The oligomer amounts of non-Tg mice in ELISA test were negligible in both age groups (supplemental Fig. S5, G and H ).
Reduction of SOD1 Levels in Human AD Patients but Not Other SODs-Finally, the levels of three isozymes of SOD and GSH were evaluated by Western blotting using the frontal lobes of AD patients and non-AD individuals ( Table 1). As the hippocampus of old human brains is likely to be susceptible to premortal ischemia (39), the frontal lobe, a cortical region known to be involved in AD, was used for analysis. The groups were matched for gender and age (non-AD, mean 80 Ϯ 1.9 years; AD, mean 84 Ϯ 4.6 years). The mean postmortem interval was balanced among the groups to reduce postmortem hypoxia. The duration of postmortem interval was neither related to the level of SOD1 detected by Western blot analysis nor to that of A␤ according to ELISA. Notably, the amounts of SOD1 protein were significantly decreased (ϳ30%) in AD cases as compared with non-AD cases, and no change in SOD2 or SOD3 levels was observed between AD and non-AD (Fig. 8, A  and B). GSH helps to protect cells from reactive oxygen species by acting as an antioxidant. The levels of GSH in the AD patients remained unchanged (Fig. 8C).

DISCUSSION
The contribution of SOD to AD pathogenesis has long been controversial. Several studies have reported the reduction of SOD in the frontal cortex of AD patients (40), whereas a slight elevation of SOD was demonstrated in the caudate nucleus of AD patients (41). Alternatively, other researchers have suggested that no changes in SOD levels are seen in AD brains (42). Recently, Ansari and Scheff (43) reported a strong correlation between oxidative damage levels (total SOD, GSH, catalase, thiobarbituric acid reactive substances, protein carbonyl, 3-nitrotyrosine, 4-hydroxynonenal, and acrolein) and the variable dementia status of subjects. They carefully screened and  DECEMBER 30, 2011 • VOLUME 286 • NUMBER 52 excluded materials damaged by premortem hypoxia to eliminate the possibility of affecting protein integrity; however, there have been no studies on the contribution of each SOD isozyme to AD pathogenesis.

SOD1-deficient AD Mice Show Early Memory Loss
In the current study we demonstrate the potential involvement of cytoplasmic SOD down-regulation in AD progression, as measured by cognitive impairment (Fig. 2), synaptic protein loss and Tau phosphorylation at Ser-396 (Fig. 3), plaque augmentation (Fig. 4), neuronal inflammation (Fig. 5), oxidative damages (Fig. 6), and the modulation of soluble A␤ state (Fig.  7). It should be highlighted that the intracellular oxidative stress in AD patients increases the Tau phosphorylation (32) and the clear concordance between Tau abnormality and cognitive impairment in animal models (36,44). Using human materials, the specific decrease of SOD1 was found in the AD patients compared with non-AD individuals.
One of the most abundant reactive oxygen species within cells (including the mitochondria) affecting memory function, synaptic plasticity (45), and neuronal death (46,47) is the superoxide radical, which implies that SOD can play a protective role in neurodegeneration. In contrast, regional alterations in oxidative damage are seen in the AD brain because of variable levels of antioxidant defenses and the consumption rate of oxygen. Even though mitochondrial oxidative stress has been hitherto thought to be one of the main contributors to AD progression (48,49), neither the levels of oligomerization nor oxidative damage in hAPP mice with haplo-deficiency of Sod2 were studied (17)(18)(19). It should be noted that SOD levels in the cytosol, nucleus, and intermembrane space of mitochondria are much higher than in the mitochondrial matrix (10). Glabe and coworkers (50) have suggested that A␤ oligomers are localized in the cytosol and membrane. Tomiyama et al. (37) implied the cytoplasmic accumulation of A␤ oligomers using their own  Shown are representative immunoblots of brain lysates using anti-CML antibody (A) and their densitometric quantification of 6 -8 months (n ϭ 4ϳ5 per genotype) (B) and 15-17 months (n ϭ 4ϳ5 per genotype) (C). D and E, the ratio of 8-oxodG to 10 7 dG using HPLC evaluation in mouse brains at the age of 6 -8 months (n ϭ 5ϳ8 per genotype) (D) and 15-17 months (n ϭ 5ϳ8 per genotype) (E) is shown. The young hAPP/Sod1 Ϫ/Ϫ revealed the most rapid increase of CML formation in proteins (ANOVA, p Ͻ 0.01). Values represent the mean Ϯ S.E.

SOD1-deficient AD Mice Show Early Memory Loss
developed mouse model with E693⌬ mutation. The recent studies by Ohyagi and co-workers (51) revealed intraneuronal A␤ accumulation and memory disturbances before extracellular deposition in unique triple transgenic AD mouse models (3ϫTg-AD). Notably, chronic hAPP overexpression decreased the activity of SOD1 in the transgenic APP23 mouse (52). Indeed, Marlatt et al. (53) implied that oxidative damage occurs primarily within the cytoplasm rather than in mitochondria.
Intriguingly, Yoon et al. (54) reported that intracellular A␤ interacts with SOD1 leading to its decreased activity. Given the direct interaction of A␤ with SOD1, it is speculated that the increased amounts of free A␤ could result in the accelerated oligomerization in hAPP/Sod1 Ϫ/Ϫ in the current study. Recently, the deletion of the copper chaperone for SOD1 increased the amounts of secreted A␤ as well as intracellular A␤ via the up-regulation of ␤APP at the ␤-site ␤APP cleaving  5 per genotype). G and H, distribution of A␤ assembly by Western blotting of 6 -8-month-old and 15-17-month-old mice with the indicated genotypes is shown. 82E1 antibody was used for A␤ detection. Overexposed bands corresponding to the ϳ30-kDa normalized to ␤-actin were used for quantification (n ϭ 4ϳ5 per genotype). Arrows represent ϳ60and ϳ30 kDa bands used for quantification below in the graph. *, p Ͻ 0.05. Values represent the mean Ϯ S.E. In the young groups, the level of only A␤ oligomers was significantly increased in hAPP/Sod1 Ϫ/Ϫ and hAPP/Sod1 ϩ/Ϫ as compared with the hAPP/Sod1 ϩ/ϩ mice. On the other hand, old hAPP/Sod1 Ϫ/Ϫ showed a significant elevation of A␤42 and A␤40 together with oligomer increase as compared with the hAPP/Sod1 ϩ/ϩ mice. enzyme (BACE1) in the study of cell cultures (55). Angeletti et al. (56) mentioned the interaction of cytoplasmic domain of BACE1 with the copper chaperone for SOD1. This led to the conclusion that cytoplasmic SOD might be one of the alternative therapeutic targets in AD.
There have been several studies on the involvement of A␤ in the drusen formation seen in AMD, a neurodegenerative disorder (57, 58). In particular, drusen deposits were reported to contain nonfibrillar amyloid oligomers (16). Previously, we have proposed Sod1 Ϫ/Ϫ mice as a valuable animal model for investigating human aging, as they showed drusen in eyes (11) as well as fatty liver (12), skin thinning (13), and osteoporosis (14). Oxidative stress was found to induce the production of A␤ in mammalian lenses (59), and Ding et al. (60) showed the inhibition of drusen pathology using anti-A␤ antibody immunotherapy. Collectively, we suggest the cytoplasmic SOD as a possible modulator of A␤ pathology and that it might function in a common mechanism shared by AD and AMD. On the other hand, as yet there is no clearly association between AMD and AD, so other processes may also need to be taken into account.
Teplow and co-workers (61) recently reported a clear relationship between A␤ oligomer size and its neurotoxicity. There have been few animal model studies on the relationship between amyloid oligomers and oxidative stress, although Klein and co-workers (62) proposed that A␤ oligomers (A␤derived diffusible ligand) induce long term potentiation accompanied with oxidative damage ex vivo. Barnham et al. (63) proposed that A␤ forms dityrosine cross-linked dimers via oxidation of the tyrosine residue at position 10 under oxidative conditions and that generic dityrosine levels were increased in the AD brain (64). We did not observe any evidence of oxidative modifications to A␤ utilizing surface-enhanced laser desorption ionization-time of flight-mass spectrometry nor any dityrosine formation in Western blotting (data not shown), implying that the promotion of oligomer formation may be the result of some as yet unidentified upstream phenomenon (e.g. endoplasmic reticulum stress, endosomal/lysosomal leakage (65)) of the intracellular superoxide radical rather than the direct interaction/modification of A␤. Further analysis will be required to clarify the intracellular oligomerization.
Selkoe and co-workers (66) also suggested that A␤ dimers are the smallest synaptotoxic species and that plaque cores are largely inactive but sequester and release dimers. Using the ELISA strategy Selkoe and coworkers (38) developed, which recognizes A␤ dimers or 2 ϫ n multimers (this implication did not contradict our data of Western blotting in Fig. 7, G and H) together with Western blotting, A␤ oligomerization in hAPP/ Sod1 Ϫ/Ϫ mice was found to be significantly elevated and corresponded to the decreased memory induced by synaptic loss (Figs. 2, 3, and 7). Quite recently, the specificity of the A␤ oligomer ELISA, especially with the use of the same antibody (6E10 raised against A␤1-16) for capture and detection, has been questioned (67). In this study the use of anti-the N-terminal A␤ antibody (82E1) may help decrease the problem using the same antibodies for capture and detection, as commented by Selkoe and co-workers (38). This approach may be improved by the combination with other methods to measure the oligomer (e.g. dot blotting using A11 oligomer antibody in Fig. 7, E and F).
It is worth noting that Sod1 reduction did not affect the amount of monomeric A␤42 or A␤40 in the 6 -8-month-old group (Fig. 7, A and C), suggesting that there may be little change in A␤ production from APP by ␤and ␥-secretases at the early stage of aging. Given the increased production of A␤42 and A␤40 at the late stage of aging (Fig. 7, B and D), . Reduction of SOD1 levels in AD subjects brain but not other SOD isozymes. A, representative immunoblots are shown for each SOD using the frontal lobe of human AD and non-AD subjects (n ϭ 10 cases per group). B, densitometric quantification of Western blot signals for each SOD isozyme normalized to ␤-actin is shown. C, levels of total GSH in the frontal lobes of AD and non-AD individuals are shown. Only SOD1 was decreased in AD brains (ANOVA, p Ͻ 0.05). Values represent the mean Ϯ S.E.
cytoplasmic SOD down-regulation might primarily affect A␤ oligomerization.
In conclusion, we have identified the in vivo involvement of cytoplasmic SOD deficiency in A␤ oligomerization and early cognitive impairment. Whether the induced oxidative stress initially affects A␤ assembly or tissue oxidation after the abnormal aggregation of A␤ still remains controversial. Alternatively, 6 -8-month aging might be a transient period of oxidative damaging against proteins or DNA. Recently, Sasaki et al. (68) reported that the superoxide-dependent chemiluminescence levels triggered by ischemia-reperfusion were significantly increased in 3-week-old Sod1 Ϫ/Ϫ . Given the recent report by Ansari and Scheff (43) that total SOD activity was not decreased in the postmitochondrial supernatant of mild cognitive impairment cases, early down-regulation of cytoplasmic SOD1 may predict the onset of dementia. The cerebral endothelial dysfunction of hAPP-overexpressing mice can be rescued by overexpression of Sod1 (69) or the administration of SOD (70). Bayer et al. (52) suggested that dietary intake of copper stabilizes SOD1 activity and attenuates A␤ production in the AD mouse model. The development of SOD1 mimetics or SOD1 stimulators from natural products may be of therapeutic benefit.